Workable

Promptfree, for everyone, and the only version there is

Copy the prompt for Claude CodeCopy the prompt for CodexCopy the prompt for Cursor

Copy

Build the hiring pipeline I actually need instead of Workable — and know what makes this ALMOST.

**Syndication is the part you cannot reproduce cheaply.** Getting a job in front of candidates means a feed to each board, most of which want their own format and an account, and several of which charge. The tracking half is a pipeline with stages, which you have built before.

STACK
- Node 20+ with Fastify, server-rendered HTML
- SQLite through better-sqlite3, WAL mode, with FTS5
- Caddy in front

THE DATA MODEL
- jobs: id, slug, title, department, location, employment_type, description_md, salary_min, salary_max, currency, is_published, opened_at, closed_at
- candidates: id, name, email, phone, cv_path, cv_text, linkedin_url, source, created_at
- applications: id, job_id, candidate_id, stage_id, answers_json, applied_at, rejected_at, rejection_reason, hired_at
- stages: id, job_id, name, position, kind — applied, screening, interview, offer, hired, rejected
- stage_history: id, application_id, from_stage_id, to_stage_id, at, actor_id — append-only, and where time-to-hire comes from. It cannot be reconstructed later
- interviews: id, application_id, starts_at_utc, ends_at_utc, kind, panel_ids_json, calendar_event_id
- scorecards: id, interview_id, interviewer_id, criteria_json, recommendation, notes, submitted_at
- comments, attachments, and an append-only activity log

THE CAREERS PAGE
- Server-rendered, fast, one page per job at a permanent URL
- **JSON-LD JobPosting on every one.** That is what puts a role into search engines' job listings, and it is free syndication that needs no account with anybody
- A feed in the common XML shapes so a board can pull rather than you pushing
- The application form is short: name, email, CV, and the two questions that actually matter. Every extra field costs applicants, measurably, and the ones you lose are the ones with other options
- Works on a phone, because most applications start on one

SCORECARDS, WHICH IS THE PART WORTH GETTING RIGHT
- Criteria defined per job before anybody interviews, not after
- Each interviewer submits independently and **cannot see the others until they have submitted.** Without that, the first opinion anchors the rest and the panel is theatre
- A recommendation on a fixed scale with a written reason, and the reason is required
- Rejection reasons from a list, because that list is the only thing that will ever tell you where the pipeline actually fails

SCHEDULING
- The same slot problem as any booking tool: availability from calendars, minus busy, minus buffers, intersected across a panel
- Every instant UTC, availability as local times with a named zone, both daylight-saving transitions tested
- The candidate picks from what fits everybody, with no account

WHAT THE LAW MAY WANT
- Equal-opportunity questions where they are required, stored **separately from the application** and never visible to anybody making a decision. That separation is the whole point of collecting them
- A retention period per application with a sweeper that deletes, and a stated basis for keeping a CV after a rejection
- Take advice: the rules differ by country and the penalties are real

MEASURING
- Time in each stage and time to hire, from stage_history
- Pass-through rate per stage, and where candidates from each source fall out
- Everything computed at query time

WHAT MATTERS MOST
JSON-LD on the job page and blind scorecards. The first gets you found without any board; the second is the difference between a hiring process and a series of opinions.